Data Preparation

Model your data as a series of process observations or measures that are associated with an outcome of interest.  Compose each observation as a common set of features (aka., independent variables or factors) along with the associated outcome.  Both features and outcomes are either numerical (dates, times, ages, etc.), binary (yes/no, true/false, etc.) or categorical (Gender, Service line , Floor unit, Shift, DRG, etc.).


Learn more »

Submission & Processing

Our best clients have these traits:

  1. They want to solve and focus on high value quality or process efficiency problems.
  2. They have process and outcome data readily or easily available through database access or other means.
  3. They have strong stakeholder focus on the promise of machine learning to improve outcomes in an ongoing basis.
  4. They are able to design interventions or strategies based, in part, on data insights associated with outcomes of interest.
  5. They are able to drive tactical execution based upon poor outcome prediction. 

What to Expect

  1. A Proof-of-Value (POV) process may take up to 4 weeks once your historical data is received and processed.
  2. We use both manual and automated machine learning to validate results and to understand how rich your dataset is in terms of predictivity.
  3. A Go/No-Go decison is made based on the POV results.
  4. Near real-time outcome prediction is an option using network transfers or applications running on your network.  Internet access is not required for predictions.


  1. We engineer features, as needed, to remove noise and improve the quality of findings and predictive models.
  2. Our machine learning engine (MLE) pairs supervised learning methods with an unsupervised learning pre-training stage to remove noise and improve predictive performance.
  3. The MLE associates patterns in the data with outcomes of interest (data mining).
  4. The MLE tests multiple supervised learning algorithms, including deep learning, to choose the most cost effective predictive model for the business goal.
  5. We use both multi-fold cross validation and separate sample testing to measure and verify the predictivity and accuracy in the predictive models.

 Automated Decision Rule Generation Provides Quick Insight

  1. Automatic generation of simple predictive decisioning rules allows you to easily distinguish factors associated with outcomes of interest, extract actionable factors and design corrective process interventions.
  2. We periodically retrain and test the predictive models to provide ongoing predictions based on fresh learning.
  3. Confidence scores accompany predictions so that interventions my be prioritized within higher risk groups.



HTO Magazine